Early release, published at www.cmaj.ca on April 8, 2020. Subject to revision.

RESEARCH HEALTH SERVICES

Mathematical modelling of COVID-19

transmission and mitigation strategies in the
population of Ontario, Canada
Ashleigh R. Tuite PhD MPH, David N. Fisman MD MPH, Amy L. Greer PhD MSc

n Cite as: CMAJ 2020. doi: 10.1503/cmaj.200476; early-released April 8, 2020

See related article at www.cmaj.ca/lookup/doi/10.1503/cmaj.200606

ABSTRACT
BACKGROUND: Physical-distancing
interventions are being used in Canada
to slow the spread of severe acute
respiratory syndrome coronavirus 2,
but it is not clear how effective they
will be. We evaluated how different
nonpharmaceutical interventions could
be used to control the coronavirus dis-
ease 2019 (COVID-19) pandemic and
reduce the burden on the health care
system.
METHODS: We used an age-structured
compartmental model of COVID-19
transmission in the population of
Ontario, Canada. We compared a base
case with limited testing, isolation and
quarantine to scenarios with the fol-
lowing: enhanced case finding, restric-
tive physical-distancing measures, or
a combination of enhanced case find-
ing and less restrictive physical dis-
tancing. Interventions were either
implemented for fixed durations or
dynamically cycled on and off, based
on projected occupancy of intensive
care unit (ICU) beds. We present medi-
ans and credible intervals from 100
replicates per scenario using a 2-year
time horizon.
RESULTS: We estimated that 56% (95%
credible interval 42%–63%) of the
Ontario population would be infected
over the course of the epidemic in the
base case. At the epidemic peak, we
projected 107 000 (95% credible interval
60 760–149 000) cases in hospital (non-
ICU) and 55 500 (95% credible interval
32 700–75 200) cases in ICU. For fixed-
duration scenarios, all interventions
were projected to delay and reduce the
height of the epidemic peak relative to
the base case, with restrictive physical
distancing estimated to have the great-
est effect. Longer duration interventions
were more effective. Dynamic interven-
tions were projected to reduce the pro-
portion of the population infected at the
end of the 2-year period and could
reduce the median number of cases in
ICU below current estimates of Ontario’s
ICU capacity.
INTERPRETATION: Without substantial
physical distancing or a combination of
moderate physical distancing with
enhanced case finding, we project that
ICU resources would be overwhelmed.
Dynamic physical distancing could
maintain health-system capacity and
also allow periodic psychological and
economic respite for populations.

T
he coronavirus disease 2019 (COVID-19) pandemic repre-
sents a global public health emergency unparalleled in
recent time. In the 2 months since the initial World
Health Organization report describing the COVID-19 outbreak
concentrated in Wuhan, China,1 the number of confirmed cases
has risen sharply from 282 to more than 330  000, with 14  510
reported deaths across all regions of the globe. 2 The first
imported case of COVID-19 in Ontario, Canada, was reported on
Jan. 25, 2020, and community transmission was first docu-
mented on Mar. 1, 2020, in British Columbia, Canada.3
This pathogen represents a substantial challenge for public
health, pandemic planning and health care systems. Severe
acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is highly
transmissible.4–7 It causes moderate to severe clinical outcomes
in about 20% of all recognized infected individuals.5,8,9 In the
absence of a vaccine, public health responses have focused on
the use of nonpharmaceutical interventions.10 These nonphar-
maceutical interventions include “case-based” measures such as
testing, contact tracing, isolation (of infected cases) and quaran-
tine (of exposed cases); and “non-case-based” measures such as
reducing the probability of transmission given an effective con-
tact (e.g., hand hygiene and cough etiquette) and physical-
distancing measures to reduce the contact rate in the popula-
tion. Physical distancing minimizes opportunities for

© 2020 Joule Inc. or its licensors CMAJ 1

 person-to-person transmission of the virus to occur. These
physical-distancing measures include some combination of
school closure, teleworking, cancellation of group activities and
RESEARCH
events, and a general overall reduction in community contacts.
Although these measures are expected to be effective in reducing
transmission of SARS-CoV-2, they are also associated with sub-
stantial economic costs and societal disruption.
Epidemiologic models can contribute important insight for
public health decision-makers by allowing for the examination
of a variety of “what-if” scenarios. The Canadian Pandemic
Influenza Plan for the Health Sector (the backbone of which
informs COVID-19 pandemic preparedness and response) iden-
tifies 2 main objectives for responding to a pandemic: to min­
imize serious morbidity and mortality, and to minimize societal
disruption.11 The overarching goal of pandemic response is to
find a combination of nonpharmaceutical interventions that
would minimize the number of cases requiring in-patient med­
ical care (e.g., hospital and intensive care unit [ICU] admis-
sions) and deaths, while also minimizing the level of societal
disruption. Societal disruption could be reduced by limiting the
overall duration that the intervention needs to be in force to
achieve the associated reductions in morbidity and mortality. A
challenge for pandemic response is that, in a fully susceptible
population, although nonpharmaceutical interventions may
slow disease transmission while they are in place, once the
intervention is lifted (or compliance with the intervention
becomes low), the transmission of the pathogen rebounds
rapid­ly.10,12 In the case of COVID-19, it may not be possible to
minimize morbidity and mortality, and societal and economic
disruption at the same time.
Given these considerations, we used a transmission dynamic
model of COVID-19 to explore the potential impact of case-based
and non-case-based nonpharmaceutical interventions in the
population of Ontario, Canada. Our analysis focuses on identify-
ing strategies that keep the number of projected severe cases
(hospital and ICU admissions) within a range that would not
overwhelm the Ontario health care system, while also consider-
ing the amount of time these interventions would be in place.
Methods
Model overview
We developed an age-structured compartmental model that
describes COVID-19 transmission in the province of Ontario, Can-
ada. We used a modified “susceptible-exposed-infectious-
recovered” framework that incorporated additional compart-
ments to account for public health interventions, different
severities of clinical symptoms and risk of hospital admission. An
overview of the model compartments and movements between
them is provided in Figure 1, and model equations and addi-
tional details are provided in Appendix 1, available at www.cmaj.
ca/lookup/suppl/doi:10.1503/cmaj.200476-/DC1. The model was
run for a period of 2 years, and we assumed that recovered indi-
viduals remained immune from re-infection for the duration of
the epidemic. Individuals remained infectious until they recov-
ered or were admitted to hospital; we did not model transmis-
2 CMAJ
sion within health care settings. For simplicity, we assumed that
all deaths occurred in cases requiring intensive care. We included
cases in hospital (non-ICU) and requiring intensive care to esti-
mate health care requirements over the course of the epidemic.
The model was constructed in R.13
Model parameters
The model was stratified by 5-year age groups using 2019 popu-
lation estimates.14 Contacts within and between age groups were
based on the POLYMOD study,15 using contact data specific for
the United Kingdom. The model was further stratified by health
status to account for differential vulnerability to severe infection
among those with underlying health conditions. We obtained
comorbidity estimates by age from the Canadian Community
Health Survey (CCHS)16 for Ontario and included the following
conditions: hypertension, heart disease, asthma, stroke, diabe-
tes and cancer. For younger age groups (<  12 yr), we used esti-
mates from Moran and colleagues.17 A limitation of the CCHS is
that it may undersample individuals from socioeconomically dis-
advantaged populations.
Parameters describing the natural history and clinical course
of infection were derived from published studies (Table 1, full
details in Appendix 1). The rate of growth of epidemics is gov-
erned by reproduction numbers, or the number of secondary
infections caused by a primary infectious case. For a pandemic
disease, in which prior immunity is absent, the operative repro-
duction number is referred to as the basic reproduction number
(R0).23 To capture variability in transmission, specifically the
observation that the basic reproduction number for COVID-19 is
overdispersed, with some cases transmitting to many others
(superspreader events), while many other cases transmit much
less, we have added volatility to the transmission term.24–26 This
causes each model run to have a different outcome owing to sto-
chasticity (i.e., random variation between model runs). The
model was initiated with 750 prevalent cases (based on 150
reported cases in Ontario on Mar. 19, 2020, and an assumed
reporting rate of 20%), that were randomly distributed across the
infectious compartments.
Interventions
Testing was assumed to move individuals with nonsevere symp-
toms from the infectious to isolated compartments. Isolated
cases were assumed to have reduced transmission compared
with nonisolated cases. Physical-distancing measures were
assumed to reduce the number of contacts per day across the
entire population. Details of parameters that were varied under
different interventions are included in Table 2. For the base case,
we assumed that there was a degree of testing and isolation
occurring and that a proportion of exposed cases were quaran-
tined. We then added in additional control measures: (i)
enhanced testing and contact tracing; (ii) restrictive physical-
distancing measures; and (iii) a combination of enhanced testing
and contract tracing, along with less restrictive physical distanc-
ing than in (ii). We considered 2 approaches to implementing
interventions: (i) fixed durations and (ii) a dynamic approach
with interventions turned on and off based on the number of
cases requiring ICU care in the population. When interventions Outputs
were not implemented, values for physical distancing and Key model outputs included final epidemic attack rates (% of
enhanced testing and contact tracing returned to base case lev- population infected at the end of the 2-year period), prevalence

els. We focused on ICU capacity, given that this is expected to be
the most limited resource during the COVID-19 epidemic. Before
the emergence of COVID-19, Ontario had about 2000 ICU beds
(0.14 beds per 1000 population), but 90% were occupied by indi-
viduals with non-COVID-19 illness. In mid-March 2020, the
Ontario government made 300 additional ventilator-associated
ICU beds available (for a total of 500 unoccupied beds (0.034 per
1000 population). As such, we used 200 COVID-19 cases in the ICU
(across all of Ontario [0.014 per 1000]) as a threshold for turning
the intervention on. This value was based on about 40% satura-
tion of available beds, combined with the recognition that there
is a lag between cases acquiring infection and requiring intensive
care, such that one would expect ICU needs to grow rapidly once
initial COVID-19 cases present for care.
RESEARCH
of hospital admissions and ICU use, and deaths. For comparison,
we show the maximum and current ICU capacity per 1000 popula-
tion relative to model projections. For the dynamic-intervention
scenarios, we also calculated the amount of time over the 2-year
model period during which the intervention was implemented,
as a measure of intervention intensity. We present model out-
puts as medians and credible intervals from 100 model replicates
per intervention; 95% credible intervals represent the range of
outcomes from the 2.5th to 97.5th percentiles, across all model
replicates.
Ethics approval
Because this study involved the use of publicly available aggre-
gate data, approval by a research ethics board was not required.

Susceptible

Exposed
Exposed
(quarantined)

Infectious
Infectious
(presymptomatic,
(presymptomatic)
isolated)

Infectious
Infectious Infectious Infectious
(mild to moderate,
(mild to moderate) (severe) (severe, isolated)
isolated)

Admitted to hospital
Admitted to hospital
(pre-ICU)

Isolated ICU

Admitted to hospital
(post-ICU)

Recovered Dead

Figure 1: Model structure of COVID-19 transmission. Exposed cases can be either quarantined or not; quarantined cases would represent those who
were identified via contact tracing. Cases admitted to hospital are assumed to be no longer infectious to others (owing to recognition of infection) and
are included in the model to estimate health care requirements. The model is stratified by age group and presence or absence of comorbidities. Note:
ICU = intensive care unit.

CMAJ 3
 Table 1: Model parameters used in the transmission model*
RESEARCH

Age group,
Parameter yr Health status Value Details Source

Latent period, d All All 2.5 Time from exposure to onset of infectiousness References 18–20
Presymptomatic infectious All All 1 Duration of infectiousness before symptom References 18–20
period, d onset
Infectious period (mild to All All 6 Symptomatic infectious period for mild-to- References 18–20
moderate), d moderate cases (in absence of isolation)
Infectious period (severe), d All All 6 Symptomatic infectious period for References 18–20
infectiousness for severe cases; assumed equal
to time to hospital admission
Basic reproduction number All All 2.3 Average number of secondary infections Reference 6
derived from a primary infection in a
susceptible population
Time in quarantine, d All All 14 Duration of quarantine for exposed cases Current policy
Relative risk of transmission All All 0.1 Isolated cases are assumed to have reduced Assumption
for cases in isolation transmission relative to unrecognized cases
Average length of stay in All All 10 Reference 21
hospital for cases not
requiring ICU care, d
Average length of stay in All All 3 For severe cases requiring ICU care Reference 21
hospital before ICU
admission, d
Average length of stay in ICU, d All All 21 For severe cases requiring ICU care Reference 22
Average length of stay in All All 21 For severe cases requiring ICU care Reference 22
hospital after ICU, d
Probability of severe infection Severe infections requiring hospital admission Reference 21
< 15 No comorbidities 0.01
15–49 No comorbidities 0.03
50–69 No comorbidities 0.12
≥ 70 No comorbidities 0.35
< 15 Comorbidities 0.02
15–49 Comorbidities 0.06
50–69 Comorbidities 0.25
≥ 70 Comorbidities 0.76
Probability severe case All All 0.26 Reference 21
requires admission to ICU
Probability of death in cases Reference 22
admitted to ICU
< 15 No comorbidities 0
15–49 No comorbidities 0.2
50–69 No comorbidities 0.36
≥ 70 No comorbidities 0.58
< 15 Comorbidities 0
15–49 Comorbidities 0.53
50–69 Comorbidities 0.9
≥ 70 Comorbidities 1

Note: ICU = intensive care unit.

*A full model description is provided in Appendix 1 (available at www.cmaj.ca/lookup/suppl/doi:10.1503/cmaj.200476-/DC1). Age group and health status refer to the population
groups to which the parameter value was applied.

4 CMAJ
 Table 2: Details of model scenarios*

RESEARCH
Scenario

Enhanced
detection with
limited physical
Enhanced case Physical distancing
Parameter Age group, yr Base case detection distancing (combination)

Nonquarantined cases tested and isolated, % < 15 10 40 10 40

15 – 49 40 60 40 60
≥ 50 70 80 70 80
Exposed cases in quarantine before infectious, % All 10 30 10 30
Reduction in contacts with physical distancing, % All 0 0 60 25
*For each scenario, the model parameters that were varied are provided above. All other parameters were as described in Table 1. When the interventions were turned off, parameter
values returned to base case values.

Results physical-distancing measures had a more modest effect, on aver-

Base case
In the model base case, with limited testing, isolation and quaran-
tine, we estimated that 56% (95% credible interval 42%–63%) of
the Ontario population would be infected over the course of the
epidemic. This would include cases of all severities. Attack rates
were projected to be highest in those aged 5–14 years (77%, 95%
credible interval 63%–83%) and 15–49 years (63%, 95% credible
interval 48%–71%). Lower attack rates were projected in individ­
uals aged younger than 5 years (50%, 95% credible interval 37%–
58%) and adults aged 50–69 years (47%, 95% credible interval
34%–55%) and 70 years and older (30%, 95% credible interval
21%–36%). An example of the outbreak trajectory across model
simulations is presented in Figure 2. At the peak of the epidemic, in
the absence of any resource constraints to provide care (i.e.,
assuming all cases requiring medical care receive it), we projected
107 000 (95% credible interval 60 760–149 000) cases in hospital
(non-ICU) and 55 500 (95% credible interval 32 700–75 200) cases in
ICU. The high prevalence of cases in ICU reflects the mean length of
age. There was substantial variability in model projections,
owing to model stochasticity.
Dynamic interventions
We also explored dynamic interventions that were turned on and
off in response to the current state of the epidemic. Dynamic
interventions were projected to be effective for reducing the pro-
portion of the population infected at the end of the 2-year
period, with potentially shorter durations of physical distancing
than the fixed-duration approach (Figure 4). For example, when
implemented dynamically, 13 months of physical distancing,
cycled on and off, reduced the median overall attack rate to 2%.
For the physical distancing alone and combination intervention
scenarios, we observed atypical epidemic curves, with the num-
ber of cases increasing and decreasing repeatedly over time. In
these scenarios, the median number of cases in ICU was reduced
below current estimates of Ontario’s ICU capacity.
Incident cases per 1000 population

ICU stay associated with COVID-19 infection in other countries.

Fixed-duration interventions
10
All of the interventions considered were projected to delay the
epidemic peak and reduce the number of cases requiring ICU care
at the peak (Figure 3). The effectiveness of the interventions
scaled with intervention duration. For all interventions, when the
5
intervention duration was 6 months or less, there was no appre-
ciable difference on final attack rate. With 12 and 18 months of
heightened response measures, the proportion of the population
infected at the end of the 2-year period was reduced, and, in some
0
simulations, the prevalence of cases requiring intensive care fell
below Ontario’s current capacity for all or part of the period. The 0 200 400 600
largest effect was observed for the restrictive physical-distancing Time, d
intervention. The combination intervention, with enhanced case
Figure 2: Projected COVID-19 epidemic trajectory for the base case model
detection and less aggressive physical distancing, was projected with minimal intervention. Daily incident cases per 1000 population are
to substantially reduce attack rates when implemented for presented. The line represents the median value of 100 model simu­
18 months, while enhanced case detection in the absence of lations, and the shaded area indicates the 95% credible interval.

CMAJ 5
 Interpretation contained,30–33 the seeding of epidemics in countries around the
globe, many with weak health systems,34 means that reintroduc-
COVID-19 poses an extraordinary challenge to societies. Whereas tion of COVID-19 will continue to occur for some time. As success-
RESEARCH

severe illness, particularly in older individuals, is frequent
enough to overwhelm a society’s ICU capacity,27 mild unrecog-
nized illness (particularly in younger individuals) contributes to
spread,28 and outbreaks may be recognized only when super-
spreader events occur,25 often in settings like health care facil­
ities.26 In contrast to severe acute respiratory syndrome (SARS),29
the high frequency of mild cases means that strategies that focus
on case identification and isolation alone are likely to fail to pre-
vent epidemic spread and overburdening of our health care sys-
tem.26 As such, population-level interventions, with their atten-
dant economic costs, have been used to prevent health systems
from collapsing.30 Although events in China, Singapore, Hong
Kong and elsewhere have shown that COVID-19 epidemics can be
ful containment efforts maintain a large number of susceptible
individuals in populations, vulnerability to repeated epidemics is
likely to persist until a COVID-19 vaccine is developed and manu-
factured at scale, or until large fractions of the population are
infected and either die or develop immunity.35
Control strategies for COVID-19 thus need to balance competing
risks: the risks of mortality and health system collapse, on the one
hand, against economic risks and attendant hardships (and health
consequences) on the other. In this work, we evaluated plausible
strategies for attenuating the COVID-19 epidemic in Ontario, Canada.
We focused on ICU resources for 2 reasons: first, because this com­
ponent of most health systems represents a scarce resource prone to
being saturated; and second, because such saturation results in

A
1 mo 3 mo 6 mo 12 mo 18 mo
4 Base case
Prevalent cases requiring ICU care

Enhanced detection of cases

Physical distancing
Combination
per 1000 population

3
ICU capacity
Current capacity
Maximum capacity
2

0
0 200 400 600 0 200 400 600 0 200 400 600 0 200 400 600 0 200 400 600
Time, d
B
Base case
Enhanced detection of cases
% of population infected

60 Physical distancing
Combination
after 2 yr

40

20

0
1 3 6 12 18
Intervention duration, mo

Figure 3: Projected intensive care unit (ICU) bed requirements and attack rates for fixed-duration interventions. (A) Prevalent cases requiring intensive
care are shown for intervention durations of 1, 3, 6, 12 and 18 months. Maximum and current ICU capacity in Ontario are indicated by the dashed
horizontal lines. Median values are presented. (B) Model-projected percentage of the population infected over the 2-year period. Attack rates include all
infections, regardless of severity. Note that the slight variability in epidemic size for the base case (with no additional intervention) reflects model
stochasticity across simulations. More extreme durations of physical distancing create the possibility of stochastic extinction (“die out”) of the disease.

6 CMAJ
abrupt surges in case-fatality, as individuals with acute respiratory added benefit of delaying the epidemic peak, which gains time that
distress syndrome will die quickly without the capacity for mechan­ can be used to build health system capacity and identify therapies
ical ventilation. In broad terms, we find that prolonged physical dis- and vaccines. However, societies remain vulnerable to resurgences

RESEARCH
tancing is the preferred strategy for maintaining ICU resources, but as long as a critical fraction of the population remains susceptible to
an extreme fixed duration of physical distancing is required to pre- disease (that fraction can be approximated as 1/R0). Evaluating how
vent the epidemic from overwhelming ICU capacity. That said, phys­ that vulnerability changes over time will require seroepidemiologic
ical distancing, even without reducing overall outbreak size, has the studies, which have not yet been performed in Canada.

A 4 B

0.3
Prevalent cases requiring ICU care

Prevalent cases requiring ICU care

3
per 1000 population

per 1000 population

Base case
0.2 Enhanced detection of cases
2 Physical distancing
Combination

ICU capacity
Current capacity
1 0.1 Maximum capacity

0 0.0
0 200 400 600 0 200 400 600
Time, d Time, d
C
60
% of population infected

40
after 2 yr

Base case
Enhanced detection of cases
Physical distancing
Combination

20

0
0 5 10 15 20
No. of months with intervention in effect

Figure 4: Projected intensive care unit (ICU) bed requirements and attack rates for dynamic interventions. (A) Prevalent cases requiring intensive care
are shown for the base case and 3 intervention scenarios. Interventions are turned on and off (returning to base-case parameter values), depending on
the number of COVID-19 cases in the ICU. Maximum and current ICU capacity in Ontario are indicated by the dashed horizontal lines. Median values are
presented. (B) Zoomed view of prevalent ICU cases to show the dynamics for the enhanced physical-distancing and combination scenarios. (C) Model-
projected estimates of percent of the population infected over the 2-year period. Attack rates include all incident infections, regardless of severity. The
amount of time the dynamic interventions are in place is shown on the x-axis. Points indicate the median duration and lines the 95% credible intervals
for each scenario.

CMAJ 7
 In contrast to fixed-duration physical distancing, we find that
dynamic physical distancing, with interventions turned on and off as
needed, based on ICU capacity crossing a given threshold,
RESEARCH
represents a more effective, and likely more palatable, control
strategy. Physical distancing can be relaxed, but this inevitably
results in resurgent disease in the population, requiring
reinstatement. Nonetheless, dynamic physical distancing is
projected to maintain ICU capacity, and dramatically reduces
overall attack rates, while requiring less total physical distancing
time than would be required by a fixed-duration strategy of
comparable effectiveness. This may be counterintuitive; however,
an important insight from our model is that dynamic interventions
can be reactivated when resurgent outbreaks are still relatively
small, leading to the high potency of such interventions.
Furthermore, dynamic physical distancing has the potential to
allow populations, and the economy, to “come up for air” at inter-
vals, which may make this strategy more sustainable. We also
found that a combination approach, with less restrictive physical
distancing along with enhanced testing, case isolation and quar-
antine, could have a similar effect in the dynamic scenario as
more restrictive physical distancing alone. It is plausible that, as
testing capacity increases, a combination approach that is less
reliant on physical distancing will strike the right balance
between disease control and societal disruption.36
Calibration to actual physical-distancing data is possible37 and
is an area for future research. However, in broad terms, less
restrictive physical-distancing regimes may be characterized by
voluntariness and allowance of small gatherings, whereas more
restrictive regimes include “lockdowns,” with individuals confined
to home and facing legal sanction for emerging without legitimate
reasons for doing so.38 Moving forward, any such disease-control
regimes need to be coupled with improved surveillance systems,
which permit needed adjustments in response to data.
Limitations
At the time of writing, well-documented limitations in testing capacity
in Ontario,39 and a lack of information on ICU occupancy by COVID-19
patients, made it challenging to know where on the epidemic curve
we currently find ourselves. The challenges in both scope of testing
and pace of testing39 make case counts a poor metric of underlying
disease activity. Acquisition of ICU occupancy data from the province
has proved challenging, though needed metrics have recently become
available from investigative reports.40 Importantly, we do not include
within-hospital transmission cycles in this iteration of our model.
Transmission in health care settings has the potential to quickly and
dramatically reduce ICU capacity, by removing trained nurses, phys­
icians and respiratory therapists from active duty, and by rapidly filling
ICUs with new patients (health care workers, and patients admitted
for other reasons with nosocomial COVID-19 infection).
Any model involves trade-offs between simplicity and realism,
and in this work we have not attempted to model physical-
distancing measures in a highly realistic way, but rather generically
as reductions in contact frequency. Our understanding of the
natural history of SARS-CoV-2 infection continues to evolve, and
the precise role of presymptomatic and subclinical transmission is
uncertain. Physical distancing becomes a more important control
8 CMAJ
measure in the face of incomplete case ascertainment owing to
asymptomatic or mildly symptomatic cases.
We do not offer precise policy prescriptions in terms of how the
reductions associated with physical distancing that we model here
are to be achieved, and we do not regard this model as a realistic
recreation of current events in Ontario. As such our model is best
interpreted qualitatively, rather than quantitatively.
The model does not include seasonality; it is possible that
transmission will attenuate in the summer,41 resulting in a decline
in cases that would be expected to resurge with the return of
colder weather. Although our model’s several limitations are a
source of uncertainty, nonetheless, the qualitative insights around
the role of physical distancing, the relatively long intervention
durations required to bend the epidemic curve, and the potential
use of cyclic interventions can be used by policy-makers and
decision-makers, along with emerging empirical evidence from
other countries, to consider the best approaches for epidemic
control over the coming months.
Lastly, we have not modelled the fact that abrupt surges in
death resulting from full ICUs would result in lower demands for
ICU beds. Our goal here is to inform policy so that such outcomes
are avoided to the extent possible.
Conclusion
We have modelled plausible contours of the COVID-19 epidemic in
Ontario, Canada, with a focus on maintenance of ICU resources. In
the absence of substantial physical distancing or a combination of
moderate physical distancing with enhanced case detection and iso-
lation, we project that ICU resources would be quickly overwhelmed,
a conclusion consistent with that in other modelling work,12 as well
as current events in Italy and Spain. On a more positive note, we
project that dynamic physical distancing, that reacts to changes in
ICU occupancy, could maintain health system capacity and also
allow periodic psychological and economic respite for populations.
References
 1. Novel coronavirus (2019-nCoV) situation report — 1: 21 January 2020. Geneva:
World Health Organization; 2020. Available: www.who.int/docs/default-source​
/coronaviruse/situation-reports/20200121-sitrep-1-2019-ncov.pdf (accessed
2020 Mar. 22).
 2. Coronavirus disease (COVID-19) situation report — 63. Geneva: World Health
Organization; 2020 Mar. 23. Available: www.who.int/docs/default-source/​
coronaviruse​/situation-reports/20200323-sitrep-63-covid-19.pdf?sfvrsn=​d97​
cb6dd_2 (accessed 2020 Mar. 24).
  3. Epidemiological data from the COVID-19 outbreak in Canada. COVID-19 Can-
ada Open Data Working Group; 2020. Available: https://github.com/ishaberry/
Covid19Canada (accessed 2020 Mar. 22).
  4. Report of the WHO–China Joint Mission on Coronavirus disease 2019 (COVID-
19). Geneva: World Health Organization; 2020 Feb. 28. Available: www.who.int/
publications​-detail/report-of-the-who-china-joint-mission-on-coronavirus-disease​
-2019-(covid-19) (accessed 2020 Mar. 22).
  5. The Novel Coronavirus Pneumonia Emergency Response Epidemiology Team.
The epidemiological characteristics of an outbreak of 2019 novel coronavirus
disease (COVID-19) — China, 2020. China CDC Weekly 2020;2:113-22.
  6. Li Q, Guan X, Wu P, et al. Early transmission dynamics in Wuhan, China, of
novel coronavirus-infected pneumonia. N Engl J Med 2020;382:1199-1207.
  7. Kucharski AJ, Russell TW, Diamond C, et al. Early dynamics of transmission and con-
trol of COVID-19: a mathematical modelling study. Lancet Infect Dis 2020; Mar 11. pii:
S1473-3099(20)30144-4. doi: 10.1016/S1473-3099(20)30144-4. [Epub ahead of print].
  8. Wu JT, Leung K, Bushman M, et al. Estimating clinical severity of COVID-19
from the transmission dynamics in Wuhan, China. Nat Med 2020 Mar. 19. doi:
10.1038/s41591-020-0822-7.
  9. Weiss P, Murdoch DR. Clinical course and mortality risk of severe COVID-19.
Lancet 2020;395:1014-5.
10. Ferguson NM, Cummings DA, Fraser C, et al. Strategies for mitigating an influ-
enza pandemic. Nature 2006;442:448-52.
11. Henry B. Canadian Pandemic Influenza Preparedness (CPIP) Task Group. Cana-
dian pandemic influenza preparedness: health sector planning guidance. Can
Commun Dis Rep 2018;44:6-9.
12. Ferguson NM, Laydon D, Nedjati-Gilani G, et al. Impact of non-pharmaceutical in-
terventions (NPIs) to reduce COVID-19 mortality and healthcare demand. London
(UK) Imperial College COVID-19 Response Team; 2020 Mar. 16. Available: www.
imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships​/
Imperial​-College-COVID19-NPI-modelling-16-03-2020.pdf (accessed 2020 Mar. 22)
13. R Core Team. R: A language and environment for statistical computing. R
Foundation for Statistical Computing, Vienna, Austria.
14. Table 17-10-0005-01. Population estimates on July 1st, by age and sex. Ottawa:
Statistics Canada; 2019.
15. Mossong J, Hens N, Jit M, et al. Social contacts and mixing patterns relevant to
the spread of infectious diseases. PLoS Med 2008;5:e74.
16. Canadian Community Health Survey, public use microdata files. Cat no
82M0013X. Ottawa: Statistics Canada; 2016. Available: www150.statcan.gc.ca/
n1/en/catalogue/82M0013X (accessed 2020 Mar. 15).
17. Moran K, Maaten S, Guttmann A, et al. Influenza vaccination rates in Ontario children:
implications for universal childhood vaccination policy. Vaccine 2009;​27:​2350-5.
18. Bi Q, Wu Y, Mei S, et al. Epidemiology and transmission of COVID-19 in Shen-
zhen China: analysis of 391 cases and 1286 of their close contacts. medRxiv
2020 Mar. 27. doi: 10.1101/2020.03.03.20028423.
19. Tindale L, Coombe M, Stockdale J, et al. Transmission interval estimates sug-
gest pre-symptomatic spread of COVID-19. medRxiv 2020 Mar. 6. doi: 10.1101/​
2020.03.03.20029983.
20. Woelfel R, Corman V, Guggemos W, et al. Clinical presentation and virological
assessment of hospitalized cases of coronavirus disease 2019 in a travel-associated
transmission cluster. medRxiv 2020 Mar. 8. [Epub ahead of print]. doi: 10.1101/​
2020.03.05.20030502.
21. Wang D, Hu B, Hu C, et al. Clinical characteristics of 138 hospitalized patients
with 2019 novel coronavirus-infected pneumonia in Wuhan, China. JAMA 2020
Feb. 7. [Epub ahead of print]. doi: 10.1001/jama.2020.1585.
22. Yang X, Yu Y, Xu J, et al. Clinical course and outcomes of critically ill patients
with SARS-CoV-2 pneumonia in Wuhan, China: a single-centered, retrospec-
tive, observational study. Lancet Respir Med 2020 Feb 24. [Epub ahead of
print]. pii: S2213-2600(20)30079-5. doi: 10.1016/S2213-2600(20)30079-5.
23. Pandemic Influenza Outbreak Research Modelling Team (Pan-infORM); Fisman D.
Modelling an influenza pandemic: a guide for the perplexed. CMAJ 2009;​181:​171-3.
24. Camacho A, Kucharski A, Aki-Sawyerr Y, et al. Temporal changes in Ebola
transmission in Sierra Leone and implications for control requirements: a real-
time modelling study. PLoS Curr 2015; Feb 10;7. pii: ecurrents.outbreaks.406ae
55e83ec0b5193e30856b9235ed2.
25. Liu Y, Eggo RM, Kucharski AJ. Secondary attack rate and superspreading
events for SARS-CoV-2. Lancet 2020;395:e47.
Competing interests: Amy Greer receives funding from the Natural Sci-
ences and Engineering Research Council of Canada, the Canadian
Institutes of Health Research (CIHR) and the Canada Research Chairs
Program. No other competing interests were declared.
This article has been peer reviewed.
Affiliations: Dalla Lana School of Public Health (Tuite, Fisman), Univer-
sity of Toronto, Ont.; Department of Population Medicine (Greer), Uni-
versity of Guelph, Guelph, Ont.
Contributors: All authors contributed to the conception and design of the
work; acquisition, analysis and interpretation of data; drafting the work
and revising it critically. All authors gave final approval to the version to be
published and agreed to be held accountable for all aspects of the work.
Funding: The research was supported by a grant to David Fisman from
CIHR (2019 COVID-19 rapid research funding OV4-170360).
Data sharing: All data used for parameterization of this model is in the
public domain and can be accessed through references cited in the
manuscript and technical appendix. Model code is not currently avail-
able, but will be made available in the coming weeks, when it is properly
CMAJ 9
26. Frieden TR, Lee CT. Identifying and interrupting superspreading events — impli-
cations for control of severe acute respiratory syndrome coronavirus 2. Emerg
Infect Dis 2020;Mar. 18;26. [Epub ahead of print] doi: 10.3201/eid2606.200495.
RESEARCH
27. Cereda D, Tirani M, Rovida F, et al. The early phase of the COVID-19 outbreak in
Lombardy, Italy. Available: https://arxiv.org/ftp/arxiv/ papers/2003/​2003​.09320.
pdf (accessed 2020 Mar. 24).
28. Li R, Pei S, Chen B, et al. Substantial undocumented infection facilitates the
rapid dissemination of novel coronavirus (SARS-CoV2). Science 2020 Mar. 16.
[Epub ahead of print]. pii: eabb3221. doi: 10.1126/science.abb3221.
29. Lipsitch M, Cohen T, Cooper B, et al. Transmission dynamics and control of
severe acute respiratory syndrome. Science 2003;300:1966-70.
30. Wang C, Liu L, Hao X, et al. Evolving epidemiology and impact of non-pharma-
ceutical interventions on the outbreak of Coronavirus disease 2019 in Wuhan,
China. medRxiv 2020 Mar. 6. doi: 10.1101/2020.03.03.20030593.
31. Koo J, Cook AR, Park M, et al. Interventions to mitigate early spread of SARS-
CoV-2 in Singapore: a modelling study. Lancet Infect Dis 2020 Mar. 23. [Epub
ahead of print]. doi: 10.1016/S1473-3099(20)30162-6.
32. Cowling BJ, Ali ST, Ng TWY, et al. Impact assessment of non-pharmaceutical
interventions against COVID-19 and influenza in Hong Kong: an observational
study. medRxiv 2020 Mar. 16. doi: 10.1101/2020.03.12.20034660.
33. Lai S, Ruktanonchai NW, Zhou L, et al. Effect of non-pharmaceutical interven-
tions for containing the COVID-19 outbreak in China. medRxiv 2020 Mar. 13.
[Epub ahead of print]. doi: 10.1101/2020.03.03.20029843.
34. Tuite AR, Bogoch, II, Sherbo R, et al. Estimation of Coronavirus disease 2019
(COVID-19) burden and potential for international dissemination of infection
from Iran. Ann Intern Med 2020 Mar. 16. doi: 10.7326/M20-0696.
35. Pandemic Influenza Outbreak Research Modelling Team. Fisman D. Modelling
an influenza pandemic: A guide for the perplexed. CMAJ 2009;181:171-3.
36. Lee VJ, Chiew CJ, Khong WX. Interrupting transmission of COVID-19: lessons
from containment efforts in Singapore. J Travel Med 2020 Mar. 13. [Epub ahead
of print]. pii: taaa039. doi: 10.1093/jtm/taaa039.
37. Thompson S, Serkez Y, Kelley L. How has your state reacted to social distancing?
The New York Times 2020 Mar. 23. Available: www.nytimes.com/interactive/​
2020/03/23/opinion/coronavirus-economy-recession.html (accessed 2020 Apr. 1).
38. Coronavirus disease 2019 (COVID-19) pandemic: increased transmission in the
EU/EEA and the UK — seventh update: 2020 Mar. 25. Sweden: European Centre
for Disease Prevention and Control; 2020. Available: www.ecdc.europa.eu/
sites/default/files/documents/RRA-seventh-update-Outbreak-of-coronavirus​
-disease-COVID-19.pdf (accessed 2020 Apr. 1).
39. Russell A, Cain P. Coronavirus: Ontario lags behind all other provinces in COVID-19
testing. Global News; updated 2020 Mar. 31. Available: https://globalnews.ca/
news/6752239/coronavirus-ontario-behind-covid-19-testing (accessed 2020 Apr. 1).
40. Crawley M. Ontario intensive care units see jump in COVID-19 patients. CTV News;
updated 2020 Mar. 28. Available: www.cbc.ca/news/canada/toronto/covid​-19​
-ontario-hospital-intensive-care-icu-patients-1.5512926 (accessed 2020 Apr. 1).
41. Kissler SM, Tedijanto C, Goldstein E, et al. Projecting the transmission dynam-
ics of SARS-CoV-2 through the post-pandemic period. medRxiv 2020 Mar. 6.
doi: 10.1101/2020.03.04.20031112.
annotated and cleaned. In the interim, those interested in model code
should contact David Fisman directly (david.fisman@utoronto.ca).
Acknowledgements: The authors thank Gabrielle Brankston, Shannon
French, Tanya Rossi and Matthew Van Camp from the Department of Popu-
lation Medicine, University of Guelph for helping to compile data on popu-
lation demographics and chronic conditions. The authors gratefully
acknowledge assistance and input from Nelson Lee (University of Alberta),
Allison McGeer (Mount Sinai Hospital), Janine McCready (Michael Garron
Hospital, Toronto), Dick Zoutman (Scarborough Hospital Network),
Jacqueline Willmore (Ottawa Public Health), Lydia Cheng (Peel Public
Health), Monali Varia (Peel Public Health), Kristen Wheeler (Halton Public
Health), Herveen Sachdeva (Toronto Public Health), Michael Finkelstein
(Toronto Public Health), Monir Taha (Ottawa Public Health), Vera Etches
(Ottawa Public Health), Isaac Bogoch (University Health Network), Chris
Kandel (University Health Network and Michael Garron Hospital), Jeff
Powis (Michael Garron Hospital) and Bart Harvey (Hamilton Public Health)
in the formulation of plausible intervention scenarios tested in this analysis.
Accepted: Apr. 2, 2020
Correspondence to: Ashleigh Tuite, ashleigh.tuite@utoronto.ca